Cross-National Moral Beliefs: The Influence of National Religious Context

Abstract

International surveys have documented wide variation in religious beliefs and practices across nations, but does this variation in the national religious context make a difference? Building on existing theory we explain why religion should have both micro and macro-level effects on morality not sanctioned by the state and why the effects of religion differ from other forms of culture. Using two international surveys and Hierarchical Linear Modeling Techniques (HLM) we sort out the effects of national context and personal beliefs on morality with and without legal underpinnings. We find that national religious context, the respondent’s age, and religious beliefs and practices are the most consistent predictors of the sexual morality index. For morality sanctioned by the state, however, the effects for personal beliefs and practices are attenuated and the effects of the national religious context are no longer significant.

A core question for sociology is: how does society shape personal beliefs and action? Empirical tests typically address this question by looking at the effects of local social networks on the attitudes and actions of the individual; research questions that can easily be addressed with local or national surveys. But the question also calls for attention to the larger society. How does the larger cultural context shape personal beliefs and action? For the discipline’s founders, this raised a series of questions about how religion contributes to the larger social order, and how the group properties of religion shape beliefs and behaviors. Weber’s Protestant Ethic thesis and Durkheim’s arguments on the integrative capacity of religion are perhaps the most familiar, but their discussions went far beyond these frequently cited propositions. The challenge, of course, has been how to measure the larger cultural context and assess the effects.

Until recently we have not had the data or the statistical methods necessary for sorting out the effects of personal religious beliefs and the contextual effects of the larger society. International surveys, however, now provide us with standardized questionnaires given to a sample of residents in multiple nations and Hierarchical Linear Modeling Techniques (HLM) allow us to sort out the effects of national context and personal beliefs. The international surveys have clearly documented the wide variation in cultural beliefs and practices across nations, including wide variation in religion. Whereas, weekly church attendance is reported by only 2 percent of the population in Denmark and 3 percent in Russia, the rates are 46 percent in the Philippines and 62 percent in Ireland. But does this variation in the religious context of nations make a difference? Does the national religious context influence the residents’ moral attitudes? If so, does it influence all residents or only those holding religious beliefs?

Building on previous theory, we explain why religion should hold a contextual effect at the national level. Our research examines how the relationship between individual religiosity and morality is shaped by the national religious context in which people are embedded. We also distinguish between morality that has state sanctions and morality that does not. Using data from two international data collections (i.e., the 1998 International Social Survey Program and the 1997 World Values Survey) and sorting out the effects of national context and personal measures using HLM techniques, we examine the micro and macro-level effects of religion and how the effects vary for morality with and without legal underpinnings.

Religion and Morality

Although morality is infused in many elements of culture, religion holds several features that are distinct. First, and perhaps most importantly, religion is based on belief in the supernatural. Most religions are organized around gods deemed to have power and authority over social action and institutions. Durkheim and Marx were confident that belief in the gods would disappear, but they recognized that until this occurred such beliefs motivated social action. Although many forms of power and authority rely on a belief in a force greater than one’s self, god(s) are viewed as transcending other forms of authority and offering explanations that attempt to make human action and purpose intelligible.¹ The greater the gods’ power, the more demands they can make. Recent work has shown that belief in a god effectively predicts a variety of behaviors and attitudes when the god is viewed as interested in and judgmental of an individual’s behavior (Greeley 1995; Stark 2001; Bader and Froese 2005; Froese and Bader 2005).

Second, unlike other cultural elements the demands of the god(s) are outlined in a systematic set of teachings, and in scores of less formal traditions. These formal guidelines are especially important when trying to understand the relationship between religion and morality. One of Weber’s key insights was that teachings vary widely across religions. Unlike Marx, Weber viewed the relationship between religious beliefs and economic action as reciprocal. Although Weber’s (1958) Protestant Ethic has been challenged, many other historical examples could be used to illustrate this reciprocal relationship. For example, the institution of slavery had a powerful impact on Christian churches, with some American denominations developing teachings to justify the institution (Ahlstrom 1975; Raboteau 1978). But Christian teachings and institutions were also instrumental in abolishing slavery in Western Europe and the United States. Recent studies (Young 2002; Stark 2003) suggest that anti-slavery movements were guided by religious convictions and worked closely with religious institutions to mobilize support. Like Weber, we are suggesting that religion helps to explain the “why”. Why do people take certain actions and how do they explain and interpret them? We argue that the guidelines of religious teachings and the integrative capacities of religious ritual will have both micro and macro level effects on personal beliefs and action.

Despite holding unique properties, however, religion is not always distinctive in the behaviors promoted or the morality taught. In fact, religious organizations go to great lengths to have their morality adopted into legal code and enforced by the state. Once adopted into the legal code, the justification for the morality relies less on religious beliefs or a supportive religious context. Thus, to the extent that religious institutions and teachings stand alone in sanctioning a behavior, religion’s influence should increase. When religion is one of many forces sanctioning a behavior, however, its effects should be reduced. As we explore the micro and macro-level influences of religion, we will also examine how these effects vary by the type of morality being studied.

National Religious Context and Morality

Because most quantitative research on morality relies on samples from a single nation, the influence of the larger cultural or national context receives little attention. Yet virtually all sociological theories point to identities, beliefs, and social structures that go beyond individuals and their local networks. Although Durkheim’s macro-level arguments assuming a unified society with a shared collective conscience have faced sharp criticism (Pope 1976; Tilly 1981; Collins 1982), there is still the recognition of a collective identity (Alexander, 1998), shared assumptions for exchanges (Coleman 1990) and a capital gained from knowing and understanding the culture (Bourdieu 1986). That is, there is a shared recognition that the larger context does matter. When applied to the religious national context, religious beliefs about morality are transmitted through social structures as well as cultural expectations.² In more religious nations, for example, religious messages about morality will be conveyed through public discourse, public institutions, legal codes, social norms, and family structures and gender roles.³ Secular and religious people alike are exposed to this national religious context.

The religious context includes both actions and beliefs. One of Durkheim’s (1951) central propositions was that areas holding high levels of religious ritual (or social integration), will have lower rates of social deviance. Although rituals integrate members into religious groups, increasing the group’s control over the individual, the process says little about the demands the group will place on personal beliefs and action. Thus, to understand the contextual effects of religion both religious beliefs and rituals must be considered. This leads to our first hypothesis: When nations have high levels of religious belief and practice, individuals will tend to hold more conservative views of morality. We are proposing that the national religious context will affect individual morality, regardless of personal religious beliefs or practice.

But the influence of religion is not confined to social context. We argue that an individual’s beliefs and involvement can also shape personal beliefs and action. Our second hypothesis offers the following prediction: To the degree that people hold strong religious beliefs and participate in religious rituals, they will hold more conservative views of morality. This is the hypothesis that has been tested by previous research. But our data allow us to test an important second prediction: The relationship will be stronger in nations with a supportive religious context. Together these three hypotheses test the micro and macro effects of religion on morality and the interactive relationship between them.

Despite being central to the arguments of early sociological theory, these three hypotheses have received limited research attention. Even when we move beyond morality, research studies testing the more general effects of religion on social control are remarkably sparse. After reviewing existing research on the relationship between religion and crime, Byron R. Johnson and colleagues (2000:46) concluded that “[r]eligion is a large part of many people’s lives, but it is not a large part of criminological research.” At the macro-level, the small group of studies using ecological units of analysis has consistently found that religion tends to reduce most forms of social deviance (Stark et al. 1983; Bainbridge 1989; Stark and Bainbridge 1996; Ellison, Burr, and McCall 1997). On the micro-level, Johnson et al.’s review and a recent meta-analysis of existing research by Colin J. Baier and Bradley R. E. Wright (2001) both conclude that religion has a deterrent effect on crime and delinquency. Yet, there are contrary findings that require explanation. Research on individual religiosity and delinquency, find that the deterrent effect of religion is reduced or eliminated when samples are drawn from areas with low levels of religious activity.

The most convincing explanations for the relationship between religion and conservative morality incorporate both macro and micro-levels of explanation. Building on Durkheim’s notion of “moral communities,” Stark and others (Stark, Kent and Doyle 1982; Welch, Tittle, and Petee 1991) have proposed that individual religiosity deters delinquency, “but only in communities where the majority of people are actively religious” (Stark 1996:165). He explains that religion’s ability to deter deviance relies on a context where religion “is accepted by the majority as a valid basis for action” (1996:164). Hence, religion reduces delinquency for samples selected in highly religious Utah, but not in California. Additional research using schools, counties, and SMSA’s as the “community” suggests that it isn’t just the level of religious activity, but also the level of religious consensus or homogeneity that deters deviance (Ellison, Burr, and McCall 1997; Regnerus 2003).

But none of this work addresses the national context. Moving beyond the local social networks and the institutions of family and school, what is the influence of the larger cultural context? Again, the research is sparse. Using 15 nations from the 1991 International Social Survey Program (ISSP) two studies offer findings consistent with the hypotheses reviewed. Jonathan Kelley and Nan Dirk De Graaf (1997) find that people living in religious nations hold more orthodox religious beliefs and Peer Scheepers, Manfred Te Grotenhuis, and Frans Van Der Slik (2002) conclude that both individual and national religiosity influence personal moral attitudes. A third study by Ted G. Jelen, John O’Donnell, and Clyde Wilcox (1993) finds that at the individual-level Catholics consistently hold negative attitudes toward legal abortion, but the contextual effects of Roman Catholicism run in the opposite direction. This research not only extends previous tests of religion’s contextual effect on morality it also supports the idea that religion has a greater influence on morality without state sanctions.

As noted earlier, the effects of religion should vary with the type of morality being measured. Whereas, some forms of morality, such as stealing, are uniformly sanctioned by the state, other forms, such as sexual behaviors, are seldom covered under legal codes. The result is that all members of a nation are held accountable for actions violating legal codes, but personal beliefs and informal cultural expectations serve as the guide for morality not sanctioned by the state. We expect religion to have a more powerful effect on moral issues not sanctioned by legal codes. This leads to our fourth hypothesis: The macro and micro-level effects of religion on conservative morality will be reduced when the moral issues are sanctioned by legal code. Conversely, when legal codes are absent and the morality in question is openly contested, the norms and sanctions of religion should hold greater sway. A handful of studies using regional samples from the United States offer tentative support for the hypothesis (Burkett and White 1974; Tittle and Welch 1983; Hadaway, Elifson, and Petersen 1984),⁴ but at least a couple other regional studies find that the effects of religion remain for even state sanctioned behavior (Grasmick, Kinsey, and Cochran, 1991; Grasmick, Bursik, and Cochran, 1991). None of this research, however, has tested for the effects of the national context.

Together these four hypotheses propose that the link between religion and morality will depend on individual and national religiosity, and secular sanctions. A supportive religious context will strengthen the relationship between personal religious involvement and morality, especially for morality not sanctioned by the state. But in nations with low levels of religious activity, the effects of individual religiosity will be reduced. To test the micro and macro-level effects of religion on morality, we will use international data and multi-level models.

Data and Methods

Data for this study are taken from the 1998 International Social Survey Program (ISSP) “Religion” module and the 1997 World Values Survey (WVS). Both the WVS and the ISSP are international surveys that permit cross-national analysis of the relation between subjective attitudes and broader institutional contexts. The WVS offers the advantage of including a greater number of countries, and ones with more cultural differences, but this also resulted in more sampling, collection, and measurement problems.⁵ The ISSP is an international consortium composed primarily of academic survey organizations located in Europe and North America (ISSP 2001). Because it is conducted in mainly Western industrialized nations, the sampling, measurement, and fieldwork of the ISSP is of a higher quality, and thus we use it to conduct our main analysis. Because the WVS offers a larger and more diverse sample of nations, we use it to verify our ISSP findings.

Each year the ISSP creates a module containing questions that are relevant to all countries and expressed in an equivalent vocabulary in all languages. Each country then takes the module into the field in conjunction with the country’s regular annual survey. Samples within each country are large and representative of adults.⁶ A total of 32 societies, 39,034 individuals, participated in the ISSP, but after we excluded ones that did not have information on key variables⁷ our sample included 29 nations and 35,356 cases. For analysis completed with the ISSP we use the recommended weight that adjusts for known differences between the sample and population.

The 1997 wave of the WVS is the third of four waves.⁸ It was designed to enable a cross-national comparison of values and norms on a wide variety of topics and to monitor changes in values and attitudes across the globe (Inglehart et al. 2000). The sample includes adults 18 and over in societies around the world.⁹ The WVS contains data from 55 societies with a total of 78,574 cases. After we excluded countries in the WVS that did not have information on key variables our sample included 46 nations and our sample size was 63,293. For analysis conducted with the WVS we use the recommended weight that adjusts for deviations in age, under sampling of the illiterate portions of the public, over sampling respondents in urban areas, and other known differences between the population and sample. This weight also adjusts for differences in the sample size of each country setting most to 1,500.

Approximately twenty percent of respondents in each dataset were missing information on variables needed in the analysis. In a preliminary analysis we ran all models with listwise deleted data, pairwise deleted data, and multiply imputed data, and found that our results were minimally affected by these different techniques for handling missing data. Since multiple imputation takes full advantage of the available data and avoids some of the bias in standard errors and test statistics that can accompany pairwise deletion, we chose to present our results using multiple imputation (Allison 2002).¹⁰ The final parameter estimates are the parameter estimated averages obtained from each regression produced through the imputation option in the statistical computing program, Hierarchical Linear Modeling (Raudenbush, Bryk, and Congdon 2005). With multiple imputation standard errors are calculated to reflect the uncertainty that is generated through simulated data.

MORALITY MEASURES

Morality refers to a set of normative standards and beliefs for guiding behavior and for proscribing inappropriate behavior. Some of these standards and beliefs are codified into laws sanctioned by the state, but many are not. Thus, we present models with two distinct measures of morality. First, we measure morality that is not uniformly sanctioned by legal codes. As we explain below, we construct an index from three variables in the ISSP that measure attitudes about premarital sex and cohabitation. The WVS did not include a measure of attitudes about sex before marriage, but the surveys do ask about a number of other issues related to sexual morality such as divorce, abortion, prostitution, and homosexuality.¹¹ Second, we measure morality that is consistently sanctioned by legal codes. Using questions from the ISSP on whether or not it is wrong to give the government false information, and from the WVS about respondents’ attitudes regarding public morality (e.g., cheating on one’s taxes and accepting bribes), we form two additional indexes. When combined with the indexes on sexuality, we have measures for morality with and without uniform state sanctions (Hypothesis 4) for the ISSP and the WVS. Table 1 offers details on the measures used and Table 2 provides descriptive statistics.

Table 1.

Description of Variables Included in the Analysis

Individual-level Variables (Individual)	ISSP Definition	ISSP Calculation/Value	WVS Definition	WVS Calculation/Value
Sexual morality	Factor scalea of three attitudinal measures on: (1) sexual relations before marriage, (2) living together without intending to get married, and (3) whether it is a good idea to live together before marriage (Cronbach alpha = .82).	Sexual relations before marriage ranges from 1=always wrong to 4=not wrong at all. This variable was reverse coded. The last two questions range from 1=strongly agree to 5=strong disagree.	Factor scalea on justification for: (1) divorce, (2) homosexuality, (3) abortion, and (4) prostitution (Cronbach alpha =..80)	Responses range from 1=never justified to 10=always justifiable. The scale was reverse coded.
Morality sanctioned by the state	Factor scalea of two attitudinal measures: (1) not reporting all of one’s income to pay less taxes and (2) giving the government incorrect information to get benefits that one is not entitled to (Cronbach alpha = .72).	Responses range from 1=not wrong to 4=seriously wrong.	Factor scalea of five attitudinal measures on: (1) cheating on taxes, (2) accepting a bribe, (3) buying stolen goods, (4) giving the government false information, and (5) avoiding a public transportation fare(Cronbach alpha =.75).	Responses range from 1=never justified to 10=always justifiable. The scale was reverse coded.
Religion important	Factor scalea of two measures: (1) belief in God and (2) description of self (Cronbach alpha = .79)	Belief in God ranges from 1=I don’t believe in God to 6=I know God really exists and I don’t have any doubts about it. Description of self ranges from 1=extremely religious to 7=extremely nonreligious, which was reverse coded and centered (mean=0)	Religion important	Response ranges from 1=very important to 4=not at all important, which was reverse coded and centered (mean=0).
Attendance	Attendance of religious services	Response ranges from 1=once a week or more to 6=never. Reverse coded and centered (mean=0).	Apart from weddings, funerals, and christenings, attendance of religious services	Response ranges from 1=more than once a week to 7=never, practically never. Reverse coded and centered (mean=0).

Control Variables	ISSP Definition	ISSP Calculation/Value	WVS Definition	WVS Calculation/Value
Married	Dummy variable	All others=0, Married=1	Same as ISSP	Same as ISSP
Female	Dummy variable	Male=0, Female=1	Same as ISSP	Same as ISSP
Age	Respondent’s age	Nine categories ranging from 1=up to 17 years to 75 years or more.	Respondent’s age	Six categories ranging from 1=18 to 24 to 65+.
Education	Respondent’s educational attainment	Responses range from 1=none, still at school to 7=university completed	Respondent’s educational attainment	Responses range from 1=no formal education to 9=university-level education with a degree.
Religion/Denomination	Thirty-seven religion/denominational categories collapsed into six.	Dummy variables where Catholic is the suppressed category. The other five categories are Protestant, no religion, no answer, other, and Orthodox.	Eleven religion/denominational categories collapsed into nine.	Dummy variables where Catholic is the suppressed category. The other eight are Protestant, Hindu Buddhist, Muslim, Orthodox, Jew, other religion, none, and all else, which included not applicable and no response.

Country-level Variables (Nation)	ISSP Definition	ISSP Calculation/Value	WVS Definition	WVS Calculation/Value
Country levels of Religiosity	Factor scale combining individual measures of attendance and religious importance. (Cronbach alpha=. 79)	Measure was aggregated to the country level and centered (mean=0).Higher numbers indicate higher country levels of religiosity.	Individual measure of religious importance.	Measure was aggregated to the country level and centered (mean=0). Higher numbers indicate higher country levels of religiosity.
Religious Concentration	Four denominational categories (Catholic, Protestant, Orthodox, and other) combined and operationalized via the Herfindahl index.	Hj = ΣS2ij, where S represents each denomination/religion divided by the total number of religious members in each country. Each country is represented with j, and i is the index of summation that runs over all the religious categories in each country.	Seven denominational categories (Catholic, Protestant, Orthodox, Jewish, Muslim, Hindu, and Buddhist) combined and operationalized via the Herfindahl index.	Hj = ΣS2ij, where S represents each denomination/religion divided by the total number of religious members in each country. Each country is represented with j, and i is the index of summation that runs over all the religious categories in each country.
Migration	Number of persons entering and leaving a country during the year per 1,000 persons (based on midyear population)	Positive numbers indicate net immigration (an excess of persons entering the country) and negative number refer to net emigration (an excess of persons leaving the country)	Same as ISSP	Same as ISSP
Corruption	Perception of corruption in a country.	Ranges from 1=least corrupt to 10=most corrupt.	Same as ISSP	Same as ISSP
Dominant Religion	Religion with the most adherents in each country.	Dummy variables where Catholic is the suppressed category. The other two categories are Protestant and Orthodox.	Same as ISSP	Dummy variables where Catholic is the suppressed category. The other five categories are Protestant Orthodox, Muslim, Hindu, and Buddhist.
Homosexuality Regulation	Not used in ISSP analysis	Not used in ISSP analysis	Whether the country has laws that discriminate against homosexuals and whether same sex unions are permitted.	0= No regulation, 1= Regulation. Questions were then added together. Higher numbers mean less tolerant.

^aAll factor scales were created using principal components’ analysis and the factor loadings were weighted in the formation of the index.

Table 2.

Frequencies of Variables Used in Analysis

Individual-level Variables	ISSP				WVS
	Mean/Proportion	SD	Min	Max	Mean/Proportion	SD	Min	Max
Sexual morality	−0.03	0.98	−2.94	3.78	−0.02	1.01	−3.79	3.6
State sanctioned morality	−0.02	0.99	−3.37	3.67	−0.03	1.02	−5.2	4.82
Religious importance	0.02	1	−3.06	2.68	2.83	1.05	1	4
Attendance	2.99	1.88	1	6	3.65	2.03	1	7
Protestant	24%	0.43	0	1	13%	0.34	0	1
None	22%	0.41	0	1	19%	0.4	0	1
No answer/else	3%	0.17	0	1	4%	0.2	0	1
Other	2%	0.15	0	1	3%	0.17	0	1
Orthodox	6%	0.23	0	1	17%	0.37	0	1
Catholic	43%	0.49	0	1	32%	0.47	0	1
Jewish					1%	0.12	0	1
Muslim					6%	0.23	0	1
Hindu	Not in ISSP				3%	0.18	0	1
Budd					1%	0.1	0	1
Female	54%	0.5	0	1	52%	0.5	0	1
Age	45.30	7.17	16	95	3.18	1.59	1	6
Degree	4.55	1.42	1	7	5.4	2.37	0	9
Married	60%	0.49	0	1	59%	0.49	0	1
Individual-level N=	35,356				63,293
Country-level Variables
Country Religiosity	0.02	0.44	−1.25	0.79	2.77	0.51	1.65	3.92
Catholic	55%	0.51	0	1	41%	0.50	0	1
Protestant	38%	0.49	0	1	24%	0.43	0	1
Orthodox	7%	0.26	0	1	22%	0.42	0	1
Muslim	Not in ISSP				7%	0.25	0	1
Hindu					2%	0.15	0	1
Buddhist					4%	0.21	0	1
Religious concentration	0.72	0.21	0.33	0.99	0.73	0.21	0.26	1
Migration	1.85	1.81	−1.23	6.07	0.19	2.61	−5.41	9.72
Corruption	4.23	2.12	1.5	8.4	5.98	2.16	1.6	9.7
Homosexuality regulation	Not in ISSP				0.96	0.87	0	2
Country-level N=	29				46

INDIVIDUAL-LEVEL MEASURES

The individual-level variables of particular interest are religious ritual participation and personal religious beliefs (Hypothesis 2 and 3). For a measure of participation in religious rituals we include religious attendance. To assess the influence of religious importance, we include a measure of personal religiosity. Previous research has suggested that there are a series of control variables closely linked to both morality and religion, which we included. These are gender (De Vaus and McAllister 1987), age (Fowler 1981), education (Darnell and Sherkat 1997; Lehrer 1999; Feigelman, Gorman, and Varacalli 1992), marital status (Mahoney, Paragament, Jewell, Swank, Scott, Emery, and Rye 1999; Mahoney, Pargament, Tarakeshwar, and Swank 2001), and respondents’ religion (D’Antonio and Cavanaugh 1983). Due to missing data, various exchange rates, and differences in costs of living, we did not include an indicator of income.

COUNTRY-LEVEL MEASURES

To estimate the influence of national context, we include measures of national religious involvement and beliefs (Hypothesis 1). For the ISSP, country-level estimates of religiosity were determined by combining the individual religious importance and attendance measures.¹² For the WVS dataset, we calculated the average country religiosity from the individual belief measure.¹³ For both surveys higher numbers indicated higher country-levels of religiosity.

The variation across nations for our religion measure was striking. For example, using the ISSP, the percentage of people in each country who consider themselves, somewhat, very or extremely religious, ranged from less than twenty percent in East Germany and Sweden to over eighty percent in the Philippines and Nigeria. Canada and Australia fall in the middle with about 45% (see Figure 1). Countries in the WVS show a similarly diverse range of views. Whereas in East Germany and Japan less than 10 percent of the people consider themselves very religious, over 90% in Nigeria consider themselves very religious. Armenia and Basque fall in the middle with about 25% of the people considering themselves very religious.¹⁴

Figure 1.

Percentage of people who consider themselves somewhat, very or extremely religious in the ISSP

Attitudes on morality also vary remarkably. Using data from the ISSP, we find that 67% of the people in Canada and Spain think that it is seriously wrong to give the government false information in exchange for benefits, but less than 35% of the people in Latvia, Poland, and Russia feel this way. Likewise, the percentage of people who think that it is always wrong to have sexual relations before marriage ranges from 30% in Chile and the United States to less than 3% in Austria and East Germany. The WVS similarly shows that while over 55% of West Germans and Swedes think that homosexuality is acceptable (over 7 on a scale ranging from 1 to 10), only 4% of the people in India and Nigeria hold this opinion. This wide variation in attitudes illustrates that there is little consensus across nations regarding the acceptability of certain moral issues. We propose that religion will be able to explain some of these important between nation differences.

We also added some additional country-level measures suggested by previous theory and research.¹⁵ Net migration was added to serve as another measure for social integration. If, as Durkheim suggests, the macro-level effects of religion are due to social integration, entering additional social integration measures should reduce or explain away the macro-level effects of religion. For both the WVS and the ISSP we computed a net migration or emigration rate from The World Factbook (2002).

The level of religious concentration was added to test for recent findings that religious consensus more effectively deters social deviance. To measure the level of religious concentration in each country we used individual denominational/religious categories to compute a concentration index, which was operationalized via the Herfindahl index. This index was initially developed to measure the degree of firm concentration in markets (see Stigler 1964).¹⁶ The index represents the probability that any two people, selected randomly from the churched population, share the same religious faith or affiliation. Because the WVS and ISSP report different religious groups, the indexes for each survey include a different set of categories.

We controlled for the dominant religion within each country using a set of dummy variables. For coding this variable, we relied on data from the World Factbook (2003) and the World Christian Encyclopedia (Barrett et al. 2001). The number of dominant religions represented by ISSP countries differs substantially from those in the WVS. The ISSP has only Catholic (N=16), Protestant (N=11), and Orthodox (N=2) nations. In contrast, the WVS includes one Hindu, two Buddhist, three Muslim, ten Orthodox, eleven Protestant, and nineteen Catholic nations or territories.¹⁷ The dominant religions remain mostly Christian, but the WVS better represents the Orthodox Christians and offers a few cases where the dominant religion is Islam, Buddhism, or Hinduism.

When explaining morality sanctioned by the state, we also controlled for the level of government corruption at the national level.¹⁸ High levels of corruption could reduce support for state sanctioned morality. Our corruption indicator is taken from Transparency International’s Annual Corruption Perceptions Index (CPI) for 2001 and 2003 (2001, 2003).¹⁹ Using 15 data sources from nine different institutions, the CPI aggregates perceptions about the extent of corruption within countries into a single index. The extent of corruption reflects the frequency of corrupt payments, the value of bribes paid, and the resulting obstacles imposed on businesses (2002: 262).

Finally, we tried to measure the extent to which some countries attempt to regulate sexuality. Although our sexual morality index for the WVS included items on homosexuality, divorce, abortion, and prostitution, our regulation measure is limited to laws regulating same-sex relations.²⁰ The indicator is a sum of whether same-sex unions are legally permitted and whether there are laws protecting against same-sex discrimination. We wanted to include a regulation measure that more broadly assessed sexual regulation. However, we were either unable to find additional measures on the governments’ attempt to regulate or we found that the policies were hopelessly intertwined with other social issues (e.g., abortion is influenced by population concerns as well as sexuality). Our index of sexual morality for the ISSP includes only measures of attitudes about heterosexual relations prior to marriage. With few nations in the ISSP sample holding formal sanctions against extramarital sex, we do not include a sexual regulation measure for this analysis.

Analytical Strategy

To simultaneously test the micro and macro-level effects of religion on morality, we use Hierarchical Linear Modeling Techniques (HLM) and the HLM software program developed by Bryk, Raudenbush, and Congdon (2005). This technique allows us to discern variation within nations (micro-level effects) from variation between them (macro-level effects). Unlike other linear models, the random coefficients produced with these models relax the assumptions on the independence of individual observations, allowing us to enter macro-level variables (e.g., county religiosity) that are aggregated from micro-level variables (e.g., personal religious beliefs) used in the equation. This technique computes slope coefficients and intercepts that are a function of the national context. Thus, individual morality will be explained by individual characteristics as well as the national context.

We begin our analysis by first estimating Maximum Likelihood estimates to assess the contribution of individual-level variables for attitudes about sexual morality using ISSP data. We take the same steps to examine state-sanctioned morality in the ISSP and sexual and state-sanctioned morality with the WVS. The individual-level model provides the average estimate of individual attitudes net of individual characteristics. Formally, the model for examining sexual morality in the ISSP with all of the individual-level measures is:

$$\begin{array}{l}{Y}_{ij}={\mathrm{\beta }}_{0\mathrm{j}}+{\mathrm{\beta }}_{1\mathrm{j}}(\mathit{Individual}\mathit{Attendance})\\ +{\mathrm{\beta }}_{2\mathrm{j}}(\mathit{Individual}\mathit{Religious}\mathit{Importance})\\ +{\mathrm{\beta }}_3(\mathit{Sex})\\ +{\mathrm{\beta }}_4(\mathit{Age})\\ +{\mathrm{\beta }}_5(\mathit{Marital}\mathit{Status})\\ +{\mathrm{\beta }}_6(\mathit{Degree})\\ +{\mathrm{\beta }}_{7-11}(\mathit{Respontent}’s\mathit{Religion}/\mathit{Denomination})+r_{ij}.\end{array}$$

The i indexes individuals and j indexes country-level influences of moral norms and religiosity. The distribution of r_ij is assumed to be random normal with a mean of 0 and variance of σ².

To explore the effects of national context on attitudes about sexual morality, we estimate the intercept and slope of the country’s influence. Formally the model for the intercept is:

$$\begin{array}{l}{\mathrm{\beta }}_{0\mathrm{j}}={\mathrm{\gamma }}_{00}+{\mathrm{\gamma }}_{01}(\mathit{Concentration})\\ +{\mathrm{\gamma }}_{02}(\mathit{Migration})\\ +{\mathrm{\gamma }}_{03-04}(\mathit{Dominant}\mathit{Religion})\\ +{\mathrm{\gamma }}_{05}(\mathit{Country}\mathit{Level}of\mathit{Religiosity})+{\mathrm{u}}_{\text{oj}}.\end{array}$$

In this model j indexes country-level influences, β_0j is the intercept term from the individual-level equation (representing individual-levels of attitudes regarding sexual morality adjusted for individual attributes), and u_oj is a country-level norm disturbance assumed to be normally distributed with a mean of 0 and variance of t_00.

Whereas the intercept will help us estimate the coefficient for the relationship between country-levels of religiosity and sexual morality, the slope will estimate the coefficient for a cross-level interaction between a country’s average level of religiosity and individual attendance and belief for estimating sexual moral attitudes. Formally, the model for the slope is:

$$\begin{array}{l}{\mathrm{\beta }}_{1\mathrm{j}}={\mathrm{\gamma }}_{10}+{\mathrm{\gamma }}_{11}(\mathit{Country}\mathit{Level}\mathit{Religiosity})+{\mathrm{u}}_{1\mathrm{j}}.\\ {\mathrm{\beta }}_{2\mathrm{j}}={\mathrm{\gamma }}_{20}+{\mathrm{\gamma }}_{21}(\mathit{Country}\mathit{Level}\mathit{Religiosity})+{\mathrm{u}}_{2\mathrm{j}}.\end{array}$$

where β_1j is the individual religious attendance coefficient for people in country j. β_2j is the individual religious importance coefficient for people in country j.U_1j refers to a country specific disturbance in the association between country-level religiosity and individual religious importance assumed to be normally distributed with a mean of 0 and variance of t₁₁. Similarly, u_2j refers to a country specific disturbance in the association between country-level religiosity and individual religious belief. Aside from dummy variables, all variables in the analysis are centered (mean=0), which means that the intercept term represents the average moral attitude for people who are assigned the suppressed category for all dummy variables and the average value on all other variables.

Results

EXPLAINING ATTITUDES ON SEXUAL MORALITY

Our estimates of attitudes about cohabitation/premarital sexual relations are shown in Table 3. The first model provides the variance estimates that can be used to compute the amount of variance explained at the individual and country-levels for the dependent variable. As expected, there is much more variance within countries (87%) than between them (13%)²¹. Nevertheless, for HLM models 13% is a reasonably high amount of variation between contextual units.

Table 3.

^a Multilevel Estimates for Religion and Other Variables on Sexual Morality (i.e., Cohabitation, Premarital Sex) Using the ISSP

	Model 1 (baseline)		Model 2 (level 1 variables)		Model 3 (level 2 variables)		Model 4 (cross-level interaction)
	B	SE	B	SE	B	SE	B	SE
Intercept	−0.04	0.07	−0.18**	0.05	−0.18**	0.06	−0.18**	0.06
Individual-level Variables
Attendance			0.15***	0.01	0.15***	0.01	0.15***	0.01
Religious importance			0.22***	0.02	0.22***	0.02	0.22***	0.01
Sex			0.00	0.01	0.00	0.01	0.00	0.01
Age			0.01***	0.00	0.01***	0.00	0.01***	0.00
Degree			−0.06***	0.01	−0.06***	0.01	−0.06***	0.01
Married			0.07***	0.02	0.07***	0.02	0.07***	0.02
a Protestant			0.15***	0.04	0.15***	0.04	0.15***	0.04
None			0.19***	0.02	0.19***	0.02	0.19***	0.02
No answer			0.15***	0.04	0.15***	0.04	0.15***	0.04
Other			0.34**	0.09	0.34**	0.09	0.34**	0.09
Orthodox			0.10**	0.03	0.10**	0.03	0.10**	0.03
Country-level Variables
Country Levels of Religiosity					0.39**	0.13	0.31*	0.12
Religious Concentration					−0.69**	0.20	−0.70**	0.20
Migration					−0.04	0.02	−0.04	0.02
a Protestant					−0.01	0.07	−0.03	0.07
Orthodox					0.13	0.09	0.13	0.09
Cross-level Interactions
Attendance X Country Levels of Religiosity							−0.03	0.03
Religion Important X Country Levels of Religiosity							0.15***	0.04
Variance Components
Individual estimate, σ2	0.90		0.55		0.55		0.55
Attendance slope, τ11			0.0049***		0.0049***		0.0049***
Religious importance slope, τ22			0.0086***		0.0086***		0.0046***
Country intercept, τ00	0.14***		0.09***		0.06***		0.06***

⁺<.1

^*<.05

^**<.01

^***<.001

^aCatholic is the comparison category.

The second model includes only the individual-level variables. Supportive of Hypothesis 2, we see that belief and attendance are both positive and significant. Thus, as religious belief and attendance increase, the rejection of cohabitation/premarital sex also increases. All of the control variables are significant except for gender. Although Catholics had more conservative attitudes than all other groups (except the “other” category) at the bivariate level, when controls were entered for worship attendance and beliefs those who identify as Protestant, Orthodox, not having a religion, or a religion not listed here were more conservative in their attitudes about premarital sexual relations.²² Likewise, older individuals were more conservative than younger ones, and people with higher educational degrees were more liberal than those with less education. These individual-level variables alone explain 38% of the total variance within the model.²³

In the third model we include both micro and macro-level variables. When entered with individual belief and attendance, the country’s level of religiosity is significant, positive and holds a standardized beta²⁴ that is larger than any other variable in the model. As a country’s overall level of religiosity increases, so also do individuals’ negative attitudes about cohabitation/premarital sexual relations. The direction of the coefficient for religious concentration is contrary to the expectations of previous research. Rather than religious homogeneity contributing to a more conservative morality, as previous work using schools and smaller regional units have found, these results suggest that living in a religiously homogeneous country appears to lead to a greater acceptance of cohabitation/premarital sexual relations. Attitudes about cohabitation and premarital sex for people residing in Protestant or Orthodox countries do not differ significantly form those in Catholic nations. In contrast to the second model, these country-level variables are accounting for an additional 33% of the variance to be explained at the country-level.²⁵ This model also offers support for our first and second hypothesis: when the level of religiosity for individuals or countries is high, so also are conservative views of morality – even with multiple controls. The individual-level variables changed minimally when the country-level variables were entered. Together the individual and country-level variables explain 41% of the total variance within the model.

In addition to examining the direct influence of a country’s level of religiosity, we also wanted to know whether country context would have even more of an influence on individuals that attend religious services regularly or hold high levels of belief, than those who do not. Thus, in the fourth model we included the interaction between religious importance and country religiosity, and also individual attendance and country religiosity. Although the latter interaction was not significant, the interaction between individual religious importance and country-levels of religiosity was. As a country’s level of religiosity increases, disapproval of cohabitation/premarital sexual relations increase even more for people with a higher level of religiosity.

To further examine this relationship we produced the predicted values on attitudes about sexual morality for a married Catholic woman living in a predominantly Catholic country who was two SD above the mean on individual religious importance. In a country that had the mean level of religiosity (mean=.02), she had a predicted value of .40 on the moral attitudes’ index, but if she resided in a country that had a level of religiosity that was two standard deviations above the mean, her predicted moral attitudes were twice as high at .93. This finding offers support for our third hypothesis, which states that the relationship between individual religious belief and traditional moral values will be stronger in nations with a supportive religious context.

Using the WVS and models similar to those used with the ISSP, we were able to replicate the majority of the findings just reported (see Table A1, Appendix). Like the results with the ISSP, the findings are supportive of hypotheses 1 and 2. Individual religious beliefs and attendance, and country-level of religiosity remain some of the strongest influences of attitudes on sexual morality compared to other variables in the model. Individuals who attended more religious services, had stronger religious beliefs, and lived in a nation supportive of religious beliefs were more likely to hold conservative views of sexual morality. Increasing age and lower levels of education were also associated with more conservative attitudes. Females were more conservative than males, and married individuals had more conservative attitudes than others. Like the ISSP findings, dominant religion was not significantly related to sexual morality. Whereas level of religious concentration was significantly related to the premarital sexual relations/cohabitation index in the ISSP analysis, it did not hold significant coefficients when explaining the WVS sexual morality index (Model 3, Table A1).

For the WVS we also included a measure on the regulation of same-sex relations. Whereas the ISSP ‘s sexual morality index included measures on heterosexual relations outside of marriage, which face few state sanctions, the WVS index included same-sex relations. The measure for same-sex regulation is positively associated with conservative attitudes about sexual morality. Although same-sex regulation is correlated with country-levels of religiosity (Pearsons Correlation=−.295), they are both significant when entered together, suggesting that they account for separate variation. Like the previous model, we looked at the interaction between individual beliefs and attendance with country-levels of religiosity (Model 4, Table A1) and did not find a significant relationship. The most robust findings, and those replicated using both the ISSP and WVS, were the micro and macro-level effects of religion, age, and, to a lesser extent, education.

EXPLAINING ATTITUDES ON MORALITY SANCTIONED BY THE STATE

The last portion of our analysis examines the influence of individual and country-levels of religiosity on morality sanctioned by the state. Specifically, we are interested in whether the effects of religion will be weaker for moral issues consistently sanctioned by legal codes versus those issues without legal underpinnings (Hypothesis 4). Using an index of two questions from the ISSP about giving the government false information and an index of four questions related to moral violations sanctioned by the state in the WVS, we tested religion’s macro and micro influence.

Our initial (single-level) estimates of attitudes about giving the government false information are shown in Table 4. Analysis of the first model gives the amount of variance that can be explained between countries versus within on the dependent variable, and 7% of it can be found between and 93% can be found within. In contrast to the previous analysis, much more of the variance on giving the government false information can be found within countries versus between them.

Table 4.

^a Multilevel Estimates for Religion and other Variables on Morality Sanctioned by the State (i.e., Giving the Government False Information) Using the ISSP

	Model 1 (baseline)		Model 2 (level 1 variables)		Model 3 (level 2 variables)		Model 4 (cross-level interaction)
	B	SE	B	SE	B	SE	B	SE
Intercept	−0.03	0.05	−0.16**	0.05	−0.13*	0.05	−0.13*	0.05
Individual-level Variables
Attendance			0.01*	0.00	0.01*	0.01	0.01*	0.01
Religious importance			0.08***	0.01	0.08***	0.01	0.09***	0.01
Sex			0.07***	0.01	0.07***	0.01	0.07***	0.01
Age			0.01***	0.00	0.01***	0.00	0.01***	0.00
Degree			0.04***	0.01	0.04***	0.01	0.04***	0.01
Married			0.07***	0.01	0.07***	0.01	0.07***	0.01
a Protestant			0.11***	0.03	0.11***	0.03	0.11***	0.03
None			0.09**	0.03	0.09**	0.03	0.09**	0.03
No answer			−0.03	0.05	−0.03	0.05	−0.03	0.05
Other			0.13**	0.05	0.13**	0.05	0.13**	0.05
Orthodox			−0.05	0.05	−0.01	0.04	−0.01	0.04
Country-level Variables
Country Levels of Religiosity					−0.01	0.06	−0.07	0.09
Religious Concentration					0.27*	0.13	0.26+	0.13
Migration					0.01	0.02	0.01	0.02
a Protestant					−0.01	0.06	−.01	0.06
Orthodox					−0.32**	0.11	−0.32**	0.11
Corruption					−0.05**	0.02	−.05**	.02
Cross-level Interactions
Attendance X Country Levels of Religiosity							0.00	0.01
Religion Important X Country Levels of Religiosity							0.03	0.03
Variance Components
Individual variance, σ2	0.91		0.88		0.88		0.88
Attendance slope, τ11			0.0002*		0.0002*		0.002*
Religious importance slope, τ22			0.0035***		0.0035***		0.0036***
Country intercept, τ00	0.07***		0.07***		0.04***		0.04***

⁺<.1

^*<.05

^**<.01

^***<.001

^aCatholic is the comparison category.

The second model includes only the individual-level variables. Individual religious beliefs and attendance are both positive and significant. Notice, however, that the coefficients are attenuated when compared to the models on sexual morality. Whereas the standardized coefficients for religious belief and attendance were .22 and .29, and towered over all control variables except age when explaining sexual morality, they drop to .08 and .02 when explaining attitudes on giving false information to the government. This supports Hypothesis 4, which states that the micro-level effects of religion will be reduced when the moral issues are sanctioned by legal codes. For this model the individual-level variables explain barely 3% of the total variance compared to the 38% explained by the same variables in our analysis of cohabitation/premarital sexual relations (Table 3, Model 2).

In the third model we examine seven contextual variables on attitudes about giving the government false information in exchange for benefits. We found that a country’s level of religiosity was not significant in explaining individual attitudes toward morality sanctioned by the state. However, individuals in nations where the dominant religion is Orthodox Christianity were more likely to approve of giving the government false information in exchange for benefits when compared to residents of Catholic majority nations. Likewise, individuals in highly corrupt nations were more likely than people living in less corrupt countries to approve of giving the government false information in exchange for benefits. The addition of these seven country-level variables increases the overall explained variation by about 3%. This last finding adds more support for our fourth hypothesis that macro effects of religion will be reduced for forms of morality that are sanctioned by legal codes versus those that are not. Indeed, in this analysis we find that country-levels of religiosity do not have any significant influence on attitudes about giving the government false information.

Finally although country-levels of religiosity did not have a significant influence on attitudes about giving the government false information in exchange for benefits, we were curious if an interaction between individual religious importance and attendance and country- levels of religiosity might have an impact. Thus, in our fourth and final model we included the interaction between individual belief and country-levels of religiosity, and also individual attendance and country-levels of religiosity. We found that neither of these interaction terms was significantly related to attitudes about giving the government false information in exchange for benefits.

When we calculated comparable models with WVS data, our results were similar (full models can be found in the appendix, Table A2). Seven percent of the variance was found between countries and 93% within. Individual religious belief and attendance were significant and positive for explaining attitudes. If one attended religious services or found religion important, one was more likely to have disapproving attitudes about violating public morals. Similar to our findings with the ISSP data, the standardized coefficients for the importance of religion (.07) and religious attendance (.02) on public morality were much smaller than they were for private morality (.16 and .12). Like our analysis with the ISSP data, we found that as age or education increased, one was more likely to disapprove of behaviors like cheating on taxes and giving the government false information. We also found that women and married people were more likely to disapprove of cheating on taxes and giving the government false information, and that country-level of corruption was significantly related to attitudes about public morality. In contrast to people living in less corrupt countries, those living in nations with higher levels of corruption were more approving of public morality violations. There were no significant differences between countries on the basis of their majority religion. Finally, we found that country-levels of religiosity were not significantly related to public morality, which again adds support to our fourth hypothesis -- religion’s macro influence will be stronger for morality that is not sanctioned by legal codes.

Discussion/Conclusion

When applied to our earlier discussion, these findings hold several implications. First, religion’s influence relies on national religious context as well as individual beliefs and rituals. The power of the social context would seem a truism for sociologists, but past research has been more attentive to explaining individual behavior with micro-level variables. With surveys focusing on a single nation, we can easily forget the national context within which these surveys are given. This is especially true in the United States where religious individualism and the separation of church and state are stressed over community and national influences. These findings remind us that national religious context continues to shape individual opinions on morality.

Second, the influence of religion cannot be reduced to social ritual or social integration. When alternative measures of social integration (e.g. migration) were added, individual religiosity remained influential and the coefficients were highly significant regardless of the type of morality being explained. Moreover the coefficients for religious importance were consistently stronger than those for religious ritual (e.g., worship attendance). Social scientists have often felt more comfortable reducing the influence of religion to ritual or economics. Their underlying fear is that to accept the effects of beliefs as real is to accept them as true. But the “truth” of religious beliefs is not the issue. W.I. Thomas and a long line of social psychologists have reminded us that when something is defined as real, it is real in its consequences (Thomas and Thomas 1928). Regardless of how erroneous religious beliefs might seem to an outsider, they can motivate believers’ actions. The distinctive features of religious beliefs and their capacity to motivate social action must be explained and acknowledged by theory.

Third, the results offer tentative support that the religious beliefs and actions of individuals interact with the religious context of their larger environment. Recent research has documented the contextual effects of schools, communities, and counties (Stark 1996; Stark and Bainbridge 1996; Regnerus 2003); here we find that national context can also interact with personal religious beliefs. The national context offers important differences from the local context. Schools and local communities highlight the importance of social networks and local institutions, the national context points to the importance of a cultural identity and consensus that goes beyond individuals and their local networks. This is not a return to Durkheim’s shared collective conscience, but it does suggest that the larger context influences an individual’s beliefs and actions.

This finding also highlights the importance of religious beliefs. Whereas, the interaction between the national religious context and individual religious beliefs was highly significant in the ISSP, the interaction with religious attendance was not. The often weaker coefficients for religious attendance and the lack of an interaction with religious context might be due in part to the limitations of attendance as a cross-national measure. Not only does the importance of attendance vary by world religion, the availability and freedom for worship attendance varies by country. Thus, attendance might be a measure of religious opportunities and affiliation with selected religions, as well as a measure of religiosity.²⁶

The effects of religious concentration are less clear. Recent work has suggested that religious concentration, or homogeneity, should lead to higher levels of moral consensus and reinforce more traditional and conservative values. But we found no evidence of such an effect at the national level. To the contrary, when explaining the sexual morality index of the ISSP, religious concentration held a negative and highly significant coefficient. Religious diversity, not concentration, contributed to explaining conservative sexual morality. We expect that the level of analysis might explain at least part of these conflicting findings. While past research was looking at the effects of local contexts (e.g., schools and SMSAs), our research is examining the national context. Religious diversity at the national level often fails to translate into a plurality at the local level. Indeed, a plurality of religions at the national level can be the result of isolated pockets of religions at the local level. Given the inconsistent findings across our two international data files, however, we are cautious in offering bold new predictions.

The results also suggest new research questions. For example, will the effects of the national context vary by groups? Sects, communes, and other demanding religious groups have long attempted to insulate their members from the influences of the larger culture. Developing tight social networks and building clear organizational boundaries they strive to support their own moral codes. To what extent does it work? Because we were not able to separate these groups from the larger sample, we were unable to address this question.

Limited by the measures available, our research focused on personal moral beliefs rather than behavior. But we expect contextual effects on behavior as well. Just as the “moral community” arguments discussed earlier have argued that local context will shape behavior, we would expect national context to do the same. Indeed, the argument that the larger context can shape social action is at the heart of most major sociological theories.

The results also raise questions at the societal level. As predicted the national and individual-level influences of religion were greater for morality that is not sanctioned by the state; but we found that the line between moralities sanctioned by the state and those that are not is a faint one indeed. Although cheating the government was uniformly sanctioned by the state and the heterosexual relations measured in the ISSP received few state sanctions, same-sex relations drew a varied response from the state. As a result, our national measures for religion and the state’s regulation of same-sex relations both held significant coefficients when explaining the WVS morality index. This suggests that it is not only whether morality is sanctioned or not sanctioned by the state that determines the influence of religion, but it is also the degree to which the morality in question is contested. Thus future work should measure the extent to which legal codes are lenient, absent or even contradict the behaviors and beliefs proscribed by a religion.

An interesting aside is that national-levels of corruption were significantly related to the respondents’ willingness to ignore this form of morality. Although not reported in the results, we found the levels of corruption and the level of defining corruption as not being wrong, to be far higher in former Soviet nations when compared to Western nations. This concurs with ethnographic research finding that during the Soviet era many laws were viewed as immoral (rather than the breaking of the laws) (Wanner 2005).

Future work should move beyond the hot button moral issues surrounding sexuality to the more mundane issues of trust in interpersonal transactions and benevolence shown to others. Alexis de Tocqueville (1945:23) argued that “[r]eligious nations are … naturally strong on the very point on which democratic nations are weak.” Because democracies offer greater freedoms, he explained, religion was needed to “impose obligations” and restrain the selfish actions of individuals. Without limiting attention to religion, how are moral obligations defined and how do they evolve? A fertile area of study is the nations moving from highly regulated social and economic environments to nations with more freedoms. Here we have the opportunity to see the new moral obligations emerge for both economic and social exchanges. Cathy Wanner’s (2005: 19) study of Ukraine concludes that in the new market economy there is a “fierce competition” over “the diverse understandings of the moral obligations the haves have to the ‘have nots’.” Will religion bear an influence on these new obligations? If so, to what extent, for which behaviors, and in which countries?

An extension of these questions is: to what extent do moral codes remain tied to the authority of the divine? Clearly many secular laws are effective without appealing to the authority of the divine. But even for secular laws that hold a cultural consensus, we found that the influence of individual religious beliefs remained. Even the most secular states have failed to remove the divine as a source for justifying moral authority. Chairman Mao Zedong eliminated most vestiges of organized religion in China during the Cultural Revolution, and replaced the old ways with his own moral teachings; but he couldn’t eliminate all beliefs in the divine. Ironically the cult of Mao soon arose, granting him divine qualities and making him an object of prayer and confession (Zuo 1991). Religious beliefs are never the sole source of moral authority, but they hold a remarkable persistence that requires explanation.

Table A1.

^a Multilevel Estimates for Religion and Other Variables on Sexual Morality (i.e., Divorce, Abortion, Homosexuality, Prostitution) Using the WVS

	Model 1 (baseline)		Model 2 (level 1 variables)		Model 3 (level 2 variables)		Model 4 (cross-level interaction)
	B	SE	B	SE	B	SE	B	SE
Intercept	−0.13+	0.07	−0.14*	0.06	−0.16*	0.06	−0.16*	0.06
Individual-level Variables			0.16***	0.01	0.16***	0.01
Religion Important			0.06***	0.01	0.06***	0.01	0.16***	0.01
Attendance							0.06***	0.01
Age			0.08***	0.01	0.08***	0.00	0.08***	0.00
Female			−0.07***	0.01	−0.07***	0.01	−0.07***	0.01
Education			−0.06***	0.00	−0.06***	0.00	−0.06***	0.00
Married			0.09***	0.01	0.09***	0.01	0.09***	0.01
a Orthodox			0.07**	0.03	0.07***	0.02	0.07***	0.02
Jewish			0.04	0.06	0.04	0.04	0.04	0.04
Muslim			0.29**	0.08	0.30***	0.02	0.30***	0.02
Hindu			0.22***	0.04	0.22***	0.03	0.22***	0.03
Buddhist			0.09+	0.05	0.10*	0.04	0.10*	0.04
Other			0.12**	0.03	0.12***	0.02	0.12***	0.02
None			−0.02	0.02	−0.02+	0.01	−0.02+	0.01
Else			0.06	0.05	0.06+	0.03	0.06+	0.03
Protestant			0.04	0.03	0.04**	0.02	0.04**	0.02
Country-level Variables
Country Levels of Religiosity					0.36***	0.09	0.43***	0.10
Religious Concentration					−0.09	0.18	−0.08	0.18
Migration					0.00	0.01	0.00	0.01
a Protestant					0.07	0.11	0.07	0.11
Muslim					−0.06	0.16	−0.05	0.16
Orthodox					0.05	0.11	0.05	0.11
Hindu					−0.12	0.27	−0.12	0.27
Buddhist					0.16	0.19	0.17	0.19
Homosexuality Regulation					0.15**	0.05	0.15**	0.05
Cross-level Interactions
Religion Important X High Country Levels of Religiosity							−0.03	0.02
Attendance X High Country Levels of Religiosity							−0.01	0.01
Variance Components
Individual estimate, σ2	0.83		0.69		0.692		0.690
Attendance slope, τ11			0.0042***		0.0043***		0.0042***
Religious importance slope, τ22			0.0014***		0.0014***		0.0014***
Country intercept, τ00	0.22***		0.16***		0.063***		0.062***

⁺<.1

^*<.05

^**<.01

^***<.001

^aCatholic is the comparison category.

Table A2.

^a Multilevel Estimates for Religion and Other Variables on Morality Sanctioned by the State (i.e., Taxes, Bribes, Stolen Goods, False Benefits) Using the WVS

	Model 1 (baseline)		Model 2 (level 1 variables)		Model 3 (level 2 variables)		Model 4 (cross-level interaction)
	B	SE	B	SE	B	SE	B	SE
Intercept	−0.02 *	0.06	−0.16**	0.04	−0.10+	0.06	−0.10+	0.06
Individual-level Variables
Religion Important			0.07***	0.01	0.07***	0.01	0.07***	0.01
Attendance			0.01+	0.01	0.01*	0.00	0.01*	0.00
Age			0.11***	0.01	0.11***	0.00	0.11***	0.00
Female			0.11***	0.01	0.11***	0.01	0.11***	0.01
Education			0.01+	0.00	0.01***	0.00	0.01***	0.00
Married			0.15***	0.01	0.15***	0.01	0.15***	0.01
a Orthodox			−0.09*	0.04	−0.09***	0.02	−0.09***	0.02
Jewish			0.04	0.04	0.04	0.05	0.04	0.05
Muslim			0.07	0.05	0.07+	0.03	0.07+	0.03
Hindu			0.13*	0.06	0.12**	0.04	0.12**	0.04
Buddhist			0.20	0.20	0.20	0.13	0.20	0.13
Other			0.06	0.05	0.06*	0.03	0.06*	0.03
None			0.02	0.03	0.02	0.01	0.02	0.01
Else			−0.19**	0.07	−0.20***	0.03	−0.20***	0.03
Protestant			0.09**	0.03	0.09***	0.02	0.09***	0.02
Country-level Variables
Country levels of Religiosity					0.00	0.10	0.01	0.10
Concentration					−0.22	0.20	−0.22	0.20
Migration					−0.02	0.01	−0.02	0.01
a Protestant					−0.12	0.11	−0.12	0.11
Muslim					−0.12	0.18	−0.12	0.18
Orthodox					−0.13	0.10	−0.13	0.10
Hindu					0.26	0.29	0.26	0.29
Buddhist					0.01	0.19	0.02	0.19
Corruption					−0.06*	0.02	−0.06*	0.02
Cross-level Interactions
Religion Important X High Country Levels of Religiosity							0.00	0.02
Attendance X High Country Levels of Religiosity							0.00	0.01
Variance Components
Individual estimate, σ2	0.95		0.89		0.89		0.89
Attendance slope, τ11			0.0025***		0.0026***		0.0027***
Religious importance slope, τ22			0.0007***		0.0005***		0.0006***
Country intercept, τ00	0.07***		0.06***		0.05***		0.06***

⁺<.1

^*<.05

^**<.01

^***<.001

^aCatholic is the comparison category.